2025/05/08

Newest at the top

2025-05-08 20:10:39 +0200kuribas(~user@ptr-17d51eo8yjf3h7nzszs.18120a2.ip6.access.telenet.be) (Remote host closed the connection)
2025-05-08 20:10:34 +0200DigitteknohippieDigit
2025-05-08 20:09:50 +0200j1n37(~j1n37@user/j1n37) (Ping timeout: 260 seconds)
2025-05-08 20:09:08 +0200j1n37-(~j1n37@user/j1n37) j1n37
2025-05-08 20:02:43 +0200 <tomsmeding> this multi-line left-hand side is just awful
2025-05-08 20:02:35 +0200 <tomsmeding> if you have layout suggestions for that D[case] equation, I'm all ears
2025-05-08 20:01:54 +0200 <tomsmeding> the neatest presentation of the code transformation, IMO, is in https://dl.acm.org/doi/pdf/10.1145/3632878 figure 2
2025-05-08 20:01:30 +0200 <tomsmeding> it's on the lambda calculus now if you look at later versions
2025-05-08 20:00:59 +0200 <tomsmeding> https://arxiv.org/pdf/2103.15776v2 section 9.2, the typesetting/layout doesn't help
2025-05-08 20:00:27 +0200Frostillicus(~Frostilli@pool-71-174-119-56.bstnma.fios.verizon.net)
2025-05-08 20:00:20 +0200 <EvanR> that's why it's past tense
2025-05-08 20:00:06 +0200 <tomsmeding> the combinators were unreadable
2025-05-08 19:59:46 +0200 <EvanR> lol
2025-05-08 19:59:43 +0200 <tomsmeding> wouldn't sell as well, though
2025-05-08 19:59:34 +0200 <EvanR> you could called it HAD combinators
2025-05-08 19:59:07 +0200manwithluck(~manwithlu@2a09:bac5:5081:2dc::49:f6) manwithluck
2025-05-08 19:58:47 +0200Frostillicus(~Frostilli@pool-71-174-119-56.bstnma.fios.verizon.net) (Read error: Connection reset by peer)
2025-05-08 19:58:41 +0200 <EvanR> good one
2025-05-08 19:58:35 +0200 <EvanR> HAD
2025-05-08 19:58:27 +0200 <tomsmeding> it HAD combinators
2025-05-08 19:58:23 +0200 <tomsmeding> but we did away with the combinators
2025-05-08 19:58:19 +0200manwithluck(~manwithlu@2a09:bac5:5081:2dc::49:f6) (Remote host closed the connection)
2025-05-08 19:58:19 +0200 <tomsmeding> there's definitely a big homomorphism, and it's definitely automatic differentiation
2025-05-08 19:58:19 +0200 <EvanR> HA
2025-05-08 19:58:18 +0200tzh(~tzh@c-76-115-131-146.hsd1.or.comcast.net) tzh
2025-05-08 19:58:00 +0200 <EvanR> so you're saying it's not even combinatory, or differentiation
2025-05-08 19:57:21 +0200 <tomsmeding> except that there aren't even any combinators any more
2025-05-08 19:57:14 +0200 <tomsmeding> combinatory homomorphic automatic differentiation
2025-05-08 19:57:05 +0200 <EvanR> EXPLAIN
2025-05-08 19:56:59 +0200 <EvanR> CHAD
2025-05-08 19:56:29 +0200 <tomsmeding> EvanR: damn, I should have replied "but not as smooth as CHAD"
2025-05-08 19:53:15 +0200 <EvanR> ok makes sense
2025-05-08 19:52:58 +0200 <tomsmeding> because Int is discrete :)
2025-05-08 19:52:51 +0200 <tomsmeding> in the sense of AD, I can differentiate a function Int -> Int just fine: its derivative is trivial, () -> ()
2025-05-08 19:52:36 +0200 <EvanR> pretty smooth as in "I'm pretty sure..."
2025-05-08 19:52:17 +0200 <tomsmeding> EvanR: the one does not imply the other
2025-05-08 19:52:06 +0200 <EvanR> differentiates most of haskell 98. So haskell 98 is differentiable, i.e. pretty smooth
2025-05-08 19:52:06 +0200 <tomsmeding> :D
2025-05-08 19:51:59 +0200 <hellwolf> efficient chad... I can run a crypto scam meme coin with this title.
2025-05-08 19:51:46 +0200 <tomsmeding> I guess the metadata is useful, yes
2025-05-08 19:51:33 +0200 <int-e> tomsmeding: the title, abstract and meta information
2025-05-08 19:51:20 +0200 <tomsmeding> int-e: what do you get from the /abs/ page?
2025-05-08 19:50:55 +0200 <int-e> . o O ( https://arxiv.org/abs/2207.03418 )
2025-05-08 19:50:51 +0200 <tomsmeding> still fairly theoretical, and there's a bit of FP fun in there too, but not as fundamental as this
2025-05-08 19:50:31 +0200 <tomsmeding> if that's a thing
2025-05-08 19:50:28 +0200 <tomsmeding> "applied FP"?
2025-05-08 19:50:03 +0200 <tomsmeding> https://arxiv.org/pdf/2207.03418 https://dl.acm.org/doi/pdf/10.1145/3632878
2025-05-08 19:49:36 +0200 <tomsmeding> also FP but writing a literature overview is unfun
2025-05-08 19:48:48 +0200 <hellwolf> but what's your phd is about, not FP?!
2025-05-08 19:48:05 +0200 <hellwolf> before that all dandy